Recent posts by fellow AEI scholars Klon Kitchen and Claude Barfield separately highlighted two important issues that must be considered together if the United States is to truly benefit from—and lead—the inevitable revolution driven by artificial intelligence technologies.
Klon Kitchen articulately laid out the argument that the Trump Administration vision for AI as a pillar of economic growth, national security, and American technological dominance is not simply a matter of taking the technological lead. Instead, the vision understands that the US must remain “the world’s center of gravity for the most powerful and transformative technology of our time” as the leading developer and deployer of the technology because “it is a general-purpose capability that will reshape industries, economies, and national power structures.”
General purpose technologies (GPTs) (such as electricity and steam engines, and including digital computing) are distinguished from other innovations for three reasons:
- their pervasiveness across diverse sectors of the economy due to their wide-ranging applicability;
- their continuous and rapid improvement, with each iteration exhibiting significant improvements in capabilities and performance; and
- their role as an innovation catalyst, serving as a platform for further inventions and innovations, enabling new possibilities across various domains.
Claude Barfield, however, drew attention to the President’s intention to impose tariffs on semiconductors manufactured abroad entering into the United States—a move he described as both foolish and dangerous. The tariff promise is in addition to the semiconductor industrial policy initiative in the CHIPS and Science Act of 2022 which provided some $39 billion to subsidize advanced semiconductor manufacturing in the US.
Semiconductors (silicon computer chips) sit at the heart of AI. Chips are designed specifically for different aspects of the vast computational effort required for AI applications. The typical approach has been for chips to be designed in countries such as the US and the UK and manufactured in countries with particular competitive advantages in their production (e.g. Taiwan, South Korea). Currently, the competition to develop and manufacture these chips is vibrant and international. Arguably, Chinese large language model (LLM) DeepSeek, a competitor to US-derived LLMs ChatGPT, Llama and Gemini, has benefited from this rivalry to produce an equally-capable version more cheaply than the originals by using chips produced in this manner.
Yet the practical implementation of AI applications will be undertaken using computing resources in the countries where the new and innovative AI-driven services are being deployed. The vast amounts of compute power to drive US applications will require chips produced at higher cost in the US (as there is no plausible way of manufacturing them at the costs observed offshore) or imported and incurring the proposed tariffs. Moreover, all of the other computing resources using chips (every cell phone, tablet, personal computer, telecommunications device and the like) will be caught in the chip tariff matter, as semiconductors are an unavoidable part of every level of general-purpose computing technology. The inevitable outcome is that every piece of equipment using semiconductors imported into the United States will be more expensive because of the tariff. This means the net cost of US computing activity will be higher than in non-tariff jurisdictions. Subsidies for local manufacturing will not help: they must be paid for by taxes in other areas, and are ultimately paid for by US citizens.
However, that is not the end of the matter. This is a fast-moving technological environment: new chips and new applications are constantly being developed. But when the equipment in which the new chips are embedded are more costly due to the tariffs, it will take longer for them to diffuse in the US than in countries where there are no tariffs. The rate at which new applications requiring the new chips are deployed will also necessarily be slower. Those who would otherwise have benefitted—US citizens—will take longer to access the benefits that will come sooner to other jurisdictions—even if the R&D for the new chips was based in the US in the first place. This is amplified by the fact that the tools that the chips enable are themselves GPTs. Every single economic sector making use of computing technology will be bound up in the deadweight of the chip tariff. New innovations in all sectors of the economy will take longer to diffuse if the newer computing power is more expensive.
The message to policy-makers is clear. The application of chip tariffs is antithetical to any vision of the United States as a leader in the application and use of AI, computing technologies and every other sector using them. A tariff will lead to higher-cost and older equipment and applications across all sectors of the economy, because the tariffed item is an essential component of a General-Purpose Technology. US leadership in far more than just the AI industry is at stake.
The post Connecting the Dots on the Chips appeared first on American Enterprise Institute – AEI.
